Extensions/ComfyUI-HiresFix-Ultra-AllInOne
ComfyUI Extension

ComfyUI-HiresFix-Ultra-AllInOne

Ultimate high-performance hires fix solution combining latent/model upscaling, sampling, and advanced histogram-based color correction into one unified node.

By ThetaCursed·Created 4 months ago·Updated 4 months ago· 29
ThetaCursed/ComfyUI-HiresFix-Ultra-AllInOne
Nodes1
On cloudLocal install
CategoryImage/Upscaling
Stars29
Updated4 months ago
Readme

ComfyUI Hires Fix Ultra - All in One

🚀 The ultimate, high-performance Hires Fix solution for ComfyUI.

<p align="center"> <img src="preview.png" alt="ComfyUI Hires Fix Ultra Node" width="600">

This node simplifies your workflow by combining latent/model upscaling, sampling, and advanced color restoration into a single, VRAM-efficient block. Say goodbye to "color washing" and complex spaghetti workflows.

✨ Key Features

  • 📦 Unified Workflow: Replaces 6+ nodes (VAE Encode, Latent Upscale, KSampler, VAE Decode, Color Match, etc.) with one clean interface.
  • 🎨 Advanced Color Correction: Features Deep Histogram Matching to eliminate the common "graying out" or "fading" effect during high-denoise upscales.
  • 🛡️ OOM Prevention (Tiled VAE): Built-in Tiled VAE support for both encoding and decoding, allowing you to upscale to 4K and beyond even on mid-range GPUs.
  • 🚀 Hybrid Upscaling: Choose between Model-based upscaling (using ESRGAN, Real-ESRGAN, etc.) or Latent-based methods (Bicubic, Bislerp, Area, etc.).
  • 📐 Smart Resizing: Automatically calculates pixel-perfect dimensions (multiples of 8) based on your upscale factor.

🛠 Installation

  1. Open your terminal and go to the ComfyUI/custom_nodes/ folder.
  2. Clone the repository:
    git clone https://github.com/ThetaCursed/ComfyUI-HiresFix-Ultra-AllInOne.git
    
  3. Restart ComfyUI.

🚀 How to Use

  1. Add the node: Image/Upscaling -> Hires Fix Ultra - All in One.
  2. Inputs:
    • Connect your base Image, Model, VAE, and Conditioning.
    • (Optional) Connect an Upscale Model if you select "Model" as the upscale method.
  3. Settings:
    • upscale_by: Set your target multiplier (e.g., 1.5x or 2.0x).
    • denoise: Set between 0.30 - 0.55 for extra detail without losing the original composition.
    • color_fix_type: Use "Deep Histogram" to maintain 1:1 color accuracy from your low-res original.

⚙️ Parameters Reference

| Parameter | Function | |-----------|----------| | upscale_method | Supports standard Latent methods (Bicubic, Bislerp) or "Model" for ESRGAN-style upscalers. | | color_fix_type | Deep Histogram: High-quality color matching. Standard: Mean/Std matching. None: Default behavior. | | tile_size_vae | Controls the Tiled VAE size. Lower this (e.g., 512) if you encounter Out of Memory errors. | | upscale_by | The scale factor. The node automatically snaps the resolution to the nearest multiple of 8. |

🧩 SEO & Search Tags

ComfyUI, Hires Fix, High-Res Fix, Stable Diffusion, Upscaling, Color Drift Fix, Tiled VAE, Latent Upscale, Image Restoration, AI Art Workflow.


If you find this node useful, please leave a ⭐ on GitHub!